Time Series Analysis

T

The_Tourist

Quote from kut2k2:
The main value of any moving average, especially adaptive moving averages as a class, is to act as lowpass filters, i.e., to separate high-frequency components of a time series from the low-frequency components. If you are a trend trader, you want to remove the high-frequency components and trade off the signals from the lowpass filtered series. OTOH if you are a RTM trader, you want to remove the low-frequency components and trade off the signals from the highpass filtered series. MAs aren't very good for RTM trading usually.
More...

Have anyone tried the application of time series forecasting models to short term future projection of ratio line for 2 instruments used in pairs/spread RTM trading? What software or excel tools are there to set up for RTM condition.i.e. which forecasting model is better for RTM? Can we account for the stationarity behaviour(deterministic) of the ratio line?
 
You could also look at using a bandpass filter. John Ehlers is the man to read for using digital signals processing in technical analysis.

However, one of the main arguments against DSP is that the market is all signal and no noise. That is for you to decide.
 
Quote from The_Tourist:

Have anyone tried the application of time series forecasting models to short term future projection of ratio line for 2 instruments used in pairs/spread RTM trading? What software or excel tools are there to set up for RTM condition.i.e. which forecasting model is better for RTM? Can we account for the stationarity behaviour(deterministic) of the ratio line?
More...

I tried this. The only thing I didn't really like about this approach is the parameterization of the model that I used. The shifts in volatility can break the model, so engineering the model was 30% of the effort, and then dealing with all the various breakdowns was 70% of the effort.

I've seen academic papers where they have some fancy algorithms that are more adaptive than AR models, and they were using some sort of filtering algorithm.

The model I used had a relatively smooth equity curve but it suffered from all the problems that your typical RTM models suffer from.
 
Quote from clearinghouse:

I tried this. The only thing I didn't really like about this approach is the parameterization of the model that I used. The shifts in volatility can break the model, so engineering the model was 30% of the effort, and then dealing with all the various breakdowns was 70% of the effort.
More...

What is your approach to deal with the breakdowns due to volatility?
 
You must log in or register to reply here.
Back
Top